Determinants of corporate credit ratings: Does ESG matter?
用十四种多分类模型评估宏观经济、财务和ESG变量对信用评级的预测能力，发现随机森林和极端随机树效果最好，且环境和社会责任变量在金融危机后影响更显著。
We perform an empirical evaluation of fourteen multinomial classifiers in the prediction of credit ratings on a large dataset consisting of macroeconomic, firm-level financial, and environmental, social, and governance (ESG) variables. Random forests and extremely randomized trees exhibit the highest predictive power for US and global firms. We show that environmental and social responsibility variables are important determinants for the credit ratings, specifically measures of environmental innovation, resource use, emissions, corporate social responsibility, and workforce determinants. The influence of ESG variables become more pronounced following the financial crisis of 2007-2009, and are important across both investment-grade and speculative-grade classes.